AI Act: Obligations for companies using ChatGPT, Copilot, and other AI tools starting August 2, 2026

Do you write emails with ChatGPT, prepare presentations with Copilot, or respond to customers with a chatbot? Then the AI Act applies to your business too. August 2, 2026 The transparency obligations of the European regulation on artificial intelligence become applicable: and they apply not only to those who develop AI, but also to those who use it every day at work.

In this guide we will see What are the obligations of the AI Act for companies?, What changes from August 2, 2026, and how to comply without turning everything into bureaucracy.

What is the AI Act and when does it come into effect?

L'AI Act (EU Regulation 2024/1689) is the world's first comprehensive law on artificial intelligence. It came into force on August 1, 2024, but will be implemented gradually:

• February 2, 2025 — bans on AI practices with unacceptable risk and AI literacy requirement (art. 4) for those who use artificial intelligence systems;
• August 2, 2025 — rules for general purpose AI models (GPAI);
• August 2, 2026 — become applicable transparency obligations (art. 50)

The key point is this: the AI Act is not just about software houses and big tech. This applies to any company that uses tools like ChatGPT, Copilot, Claude, or Gemini in their processes. —from writing emails to customer support to data analysis.

Using AI in the workplace is no longer just a productivity choice: it's also an organizational responsibility.

AI literacy: AI training is already mandatory

The aspect that surprised companies the most is the so-called AI literacy, artificial intelligence literacy. And here, pay attention to the dates: this obligation It has already been in force since February 2, 2025.

ATTENTION: Article 4 of the AI Act requires organizations using AI systems to ensure that the people involved have an adequate level of competence in the use of these tools..

It's not just about tech companies and developers: it's about every company where employees use AI to do their jobs.. If your team uses ChatGPT, Copilot, Geminio, or Claude, your organization needs to ensure they understand:

• opportunities and advantages of AI tools;
• model limitations (errors, hallucinations, biases);
• risks usage-related;
• aspects ethical and regulatory;
• good practices of security and privacy.

The goal is not to turn everyone into an AI expert, but to ensure a conscious and responsible use of technology.

There is no standard course that is valid for everyone

A common mistake is to think that a single mandatory course, the same for all employees, is enough. The AI Act doesn't work that way: does not provide standardized training.

The training must be:

• proportionate to the business context;
• consistent with the role of the person;
• appropriate to the level of use of AI.

A concrete example: those who use ChatGPT to write emails or summarize documents have very different training needs from those who use AI for select candidates, evaluate CVs or make decisions that affect people — activities classified as high-risk by the AI Act, whose specific obligations will, however, come into force later, between the end of 2027 and 2028.

This is why training must be designed on real activities carried out in the company and provided by professionals such as those at BigFive. There is no mandatory AI course that is the same for everyone and it is not enough to purchase a generic course. There is the tailor-made AI training right for your business.

Transparency: When you use AI, you have to say so.

One of the fundamental principles of the AI Act is the transparency. From August 2, 2026, when a person interacts with an AI system, they must be able to know it.

In practice, the obligation concerns very common situations:

• chatbot on the company website;
• virtual assistants for customer care;
• automatic systems responsive to customers;
• AI agents that support sales or technical support.

If the answer comes from an artificial intelligence, the user must be clearly informed about it.

The same principle applies to many AI-generated contentImages, videos, documents, and multimedia content. The goal isn't to ban them, but to prevent artificial content from being presented as entirely human—think deepfakes or synthetic content used to inform the public.

Have you read headlines like "AI Act postponed"? There is a postponement, but it only concerns high-risk systems: with the Digital Omnibus package, the deadlines are postponed to 2 December 2027 and 2 August 2028. The European Parliament approved the text on 26 March 2026, but negotiations with the Council are still ongoing.

The transparency obligations remain confirmed until August 2, 2026The only exception is an extra window, until December 2, 2026, for systems already on the market before August, limited to the technical aspects of content marking (Parliament and the Commission are not yet aligned on this date).

To guide practical application, the European Commission published the draft of the guidelines on Article 50, along with the second draft of the Code of Practice on Marking AI-generated content: two useful references for those who need to adapt chatbots and content.

How to Comply: A Checklist for Your Business

Where to start? An honest internal audit is the first step. Ask yourself:

1. Are employees using AI tools? Often the answer is yes, even when the company doesn't know it: this is the phenomenon of “shadow AI.”.
2. There are company guidelines on the use of these tools?
3. Specific training has been organised, documented and appropriate to the roles?
4. Rules have been defined for managing sensitive data and information in AI tools?
5. Chatbots and audience-facing AI content do they claim to be such?

If you answered no to one or more questions, it's time to act: the August 2, 2026, deadline is approaching, and the training requirement is already in effect.

The upside? Companies taking action now aren't just avoiding regulatory risk: they're building the skills they need to use AI effectively anyway. If you want to understand where to start with concrete applications, read our guide on’artificial intelligence for SMEs.

Conclusion: obligation or opportunity?

For the first time in Europe, understanding a technology is not only considered a useful skill, but a organizational responsibility. The AI Act asks companies to do what they should do anyway: use artificial intelligence methodically, with awareness, and with clear rules.

Those who approach it as a purely bureaucratic requirement will incur a cost. Those who approach it as an opportunity to train people and adopt AI solutions well designed will result in productivity, competitiveness and less risk operational and reputational.

Compliance with the AI Act can't be bought: it's built by training people.